1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317
|
/* -*- Mode: C; c-basic-offset:4 ; indent-tabs-mode:nil -*- */
/*
* Copyright (c) 2004-2007 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2005 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
* University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved.
* Copyright (c) 2006-2010 QLogic Corporation. All rights reserved.
* Copyright (c) 2012-2017 Los Alamos National Security, LLC. All rights
* reserved.
* Copyright (c) 2013-2017 Intel, Inc. All rights reserved
* Copyright (c) 2018 Research Organization for Information Science
* and Technology (RIST). All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
#include "ompi_config.h"
#include "opal/mca/event/event.h"
#include "opal/util/output.h"
#include "opal/util/show_help.h"
#include "opal/util/opal_environ.h"
#include "ompi/proc/proc.h"
#include "mtl_psm2.h"
#include "mtl_psm2_types.h"
#include "mtl_psm2_request.h"
#include "psm2.h"
#include <sys/types.h>
#include <sys/stat.h>
#include <unistd.h>
#include <glob.h>
static int param_priority;
/* MPI_THREAD_MULTIPLE_SUPPORT */
opal_mutex_t mtl_psm2_mq_mutex = OPAL_MUTEX_STATIC_INIT;
#if OPAL_CUDA_SUPPORT
static bool cuda_envvar_set = false;
#endif
static int ompi_mtl_psm2_component_open(void);
static int ompi_mtl_psm2_component_close(void);
static int ompi_mtl_psm2_component_query(mca_base_module_t **module, int *priority);
static int ompi_mtl_psm2_component_register(void);
static mca_mtl_base_module_t* ompi_mtl_psm2_component_init( bool enable_progress_threads,
bool enable_mpi_threads );
mca_mtl_psm2_component_t mca_mtl_psm2_component = {
{
/* First, the mca_base_component_t struct containing meta
* information about the component itself */
.mtl_version = {
MCA_MTL_BASE_VERSION_2_0_0,
.mca_component_name = "psm2",
MCA_BASE_MAKE_VERSION(component, OMPI_MAJOR_VERSION, OMPI_MINOR_VERSION,
OMPI_RELEASE_VERSION),
.mca_open_component = ompi_mtl_psm2_component_open,
.mca_close_component = ompi_mtl_psm2_component_close,
.mca_query_component = ompi_mtl_psm2_component_query,
.mca_register_component_params = ompi_mtl_psm2_component_register,
},
.mtl_data = {
/* The component is not checkpoint ready */
MCA_BASE_METADATA_PARAM_NONE
},
.mtl_init = ompi_mtl_psm2_component_init,
}
};
static int
get_num_total_procs(int *out_ntp)
{
*out_ntp = (int)ompi_process_info.num_procs;
return OMPI_SUCCESS;
}
static int
get_num_local_procs(int *out_nlp)
{
/* num_local_peers does not include us in
* its calculation, so adjust for that */
*out_nlp = (int)(1 + ompi_process_info.num_local_peers);
return OMPI_SUCCESS;
}
static int
ompi_mtl_psm2_component_register(void)
{
int num_local_procs, num_total_procs;
ompi_mtl_psm2.connect_timeout = 180;
(void) mca_base_component_var_register(&mca_mtl_psm2_component.super.mtl_version,
"connect_timeout",
"PSM2 connection timeout value in seconds",
MCA_BASE_VAR_TYPE_INT, NULL, 0, 0,
OPAL_INFO_LVL_9,
MCA_BASE_VAR_SCOPE_READONLY,
&ompi_mtl_psm2.connect_timeout);
(void) get_num_local_procs(&num_local_procs);
(void) get_num_total_procs(&num_total_procs);
/* set priority high enough to beat ob1's default (also set higher than psm) */
if ((num_local_procs == num_total_procs) && (1 < num_total_procs)) {
/* Disable hfi if all processes are local. However, if running only one
* process assume it is ompi_info or this is most likely going to spawn, for
* which all PSM2 devices are needed */
setenv("PSM2_DEVICES", "self,shm", 0);
}
param_priority = 40;
(void) mca_base_component_var_register (&mca_mtl_psm2_component.super.mtl_version,
"priority", "Priority of the PSM2 MTL component",
MCA_BASE_VAR_TYPE_INT, NULL, 0, 0,
OPAL_INFO_LVL_9,
MCA_BASE_VAR_SCOPE_READONLY,
¶m_priority);
ompi_mtl_psm2_register_pvars();
return OMPI_SUCCESS;
}
static int
ompi_mtl_psm2_component_open(void)
{
int res;
glob_t globbuf = {0};
/* Component available only if Omni-Path hardware is present */
res = glob("/dev/hfi1_[0-9]", GLOB_DOOFFS, NULL, &globbuf);
if (globbuf.gl_pathc > 0 || GLOB_NOMATCH==res) {
globfree(&globbuf);
}
if (0 != res) {
res = glob("/dev/hfi1_[0-9][0-9]", GLOB_APPEND, NULL, &globbuf);
if (globbuf.gl_pathc > 0) {
globfree(&globbuf);
}
if (0 != res) {
return OPAL_ERR_NOT_AVAILABLE;
}
}
/* Component available only if at least one hfi1 port is ACTIVE */
bool foundOnlineHfi1Port = false;
size_t i;
char portState[128];
FILE *devFile;
if (glob("/sys/class/infiniband/hfi1_*/ports/*/state",
GLOB_DOOFFS, NULL, &globbuf) != 0) {
return OPAL_ERR_NOT_AVAILABLE;
}
for (i=0;i < globbuf.gl_pathc; i++) {
devFile = fopen(globbuf.gl_pathv[i], "r");
fgets(portState, sizeof(portState), devFile);
fclose(devFile);
if (strstr(portState, "ACTIVE") != NULL) {
/* Found at least one ACTIVE port */
foundOnlineHfi1Port = true;
break;
}
}
globfree(&globbuf);
if (!foundOnlineHfi1Port) {
return OPAL_ERR_NOT_AVAILABLE;
}
return OMPI_SUCCESS;
}
static int
ompi_mtl_psm2_component_query(mca_base_module_t **module, int *priority)
{
/*
* Mixing the PSM2 MTL with the OFI BTL (using PSM2)
* can cause an issue when they both call psm2_finalize
* in older versions of libpsm2.
*/
if (!psm2_get_capability_mask(PSM2_LIB_REFCOUNT_CAP)) {
opal_output_verbose(2, ompi_mtl_base_framework.framework_output,
"This version of the PSM2 MTL needs version 11.2.173 or later of the libpsm2 library for correct operation.\n");
return OMPI_ERR_FATAL;
}
/*
* if we get here it means that PSM2 is available so give high priority
*/
*priority = param_priority;
*module = (mca_base_module_t *)&ompi_mtl_psm2.super;
return OMPI_SUCCESS;
}
static int
ompi_mtl_psm2_component_close(void)
{
#if OPAL_CUDA_SUPPORT
if (cuda_envvar_set) {
opal_unsetenv("PSM2_CUDA", &environ);
}
#endif
return OMPI_SUCCESS;
}
static int
get_local_rank(int *out_rank)
{
ompi_node_rank_t my_node_rank;
*out_rank = 0;
if (OMPI_NODE_RANK_INVALID == (my_node_rank =
ompi_process_info.my_node_rank)) {
return OMPI_ERROR;
}
*out_rank = (int)my_node_rank;
return OMPI_SUCCESS;
}
static mca_mtl_base_module_t *
ompi_mtl_psm2_component_init(bool enable_progress_threads,
bool enable_mpi_threads)
{
psm2_error_t err;
int verno_major = PSM2_VERNO_MAJOR;
int verno_minor = PSM2_VERNO_MINOR;
int local_rank = -1, num_local_procs = 0;
#if OPAL_CUDA_SUPPORT
int ret;
char *cuda_env;
glob_t globbuf = {0};
#endif
/* Compute the total number of processes on this host and our local rank
* on that node. We need to provide PSM2 with these values so it can
* allocate hardware contexts appropriately across processes.
*/
if (OMPI_SUCCESS != get_num_local_procs(&num_local_procs)) {
opal_output(0, "Cannot determine number of local processes. "
"Cannot continue.\n");
return NULL;
}
if (OMPI_SUCCESS != get_local_rank(&local_rank)) {
opal_output(0, "Cannot determine local rank. Cannot continue.\n");
return NULL;
}
err = psm2_error_register_handler(NULL /* no ep */,
PSM2_ERRHANDLER_NOP);
if (err) {
opal_output(0, "Error in psm2_error_register_handler (error %s)\n",
psm2_error_get_string(err));
return NULL;
}
#if OPAL_CUDA_SUPPORT
/*
* If using CUDA enabled Open MPI, the user likely intends to
* run with CUDA buffers. So, force-set the envvar here if user failed
* to set it.
*/
ret = glob("/sys/module/nvidia", GLOB_DOOFFS, NULL, &globbuf);
if (globbuf.gl_pathc > 0) {
globfree(&globbuf);
}
cuda_env = getenv("PSM2_CUDA");
if (!cuda_env && (0 == ret)) {
opal_show_help("help-mtl-psm2.txt",
"no psm2 cuda env", true,
ompi_process_info.nodename);
opal_setenv("PSM2_CUDA", "1", false, &environ);
cuda_envvar_set = true;
}
#endif
err = psm2_init(&verno_major, &verno_minor);
if (err) {
opal_show_help("help-mtl-psm2.txt",
"psm2 init", true,
psm2_error_get_string(err));
return NULL;
}
/* Complete PSM2 initialization */
ompi_mtl_psm2_module_init(local_rank, num_local_procs);
ompi_mtl_psm2.super.mtl_request_size =
sizeof(mca_mtl_psm2_request_t) -
sizeof(struct mca_mtl_request_t);
return &ompi_mtl_psm2.super;
}
|